GateWeekPassing criteriaRemediation path
G1 6 Course repo · weekly tracker · Dockerized service · diagnostic report Remediate orientation and systems setup
G2 18 Explain eigenvalues, SVD, norms, condition numbers; implement basic numerical routines Remediate linear/numerical algebra
G3 30 Derive basic gradients; explain MLE/MAP; implement hypothesis tests + bootstrap CIs; explain power/MDE Remediate calculus/statistics; publish Foundations release
G4 45 Completed all engineering primitives (retrieval, rate limiting, caching, chunking, PII, metrics, resilience, sketches, routing) Remediate the missing primitive contracts/tests
G5 57 Completed system design case studies with requirements, API, data model, scaling, trade-offs, failure modes Remediate the weakest design pattern
G6 69 A/B analysis tool, bootstrap tool, SRM detector, feature flag engine, experiment assignment, monitoring plan Remediate experimentation/MLOps; publish Engineering + Systems release
G7 81 RAG retrieval pipeline, retrieval evaluation report, prompt system, memory system, agent prototype Remediate the weakest LLM/RAG component
G8 93 Deployed vLLM or SGLang, benchmark methodology, observability dashboard, quantization report, speculative decoding report, inference cost model Remediate inference benchmarking and operations
G9 102 Working production AI platform: router, quotas, tracing, logging, idempotency, work queue, model registry, cost attribution, safety filters Remediate the missing platform capability
G10 108 Completed capstone, published portfolio, passed mock interviews, behavioral stories, 30-day job-search plan Remediate the specific capstone/interview gap
Portfolio releases

Three evidence-backed checkpoints

W1–30Foundations

Math, numerical routines, and experimentation toolkit

badge verified · Week 30
W31–69Engineering + Systems

Tested primitives, system designs, MLOps tools, and postmortem

badge verified · Week 69
W70–108LLM Platform

LLM/RAG work, inference benchmark, platform layer, and capstone

badge verified · Week 108
Completion is evidence-based, not attendance-based

At every level, completion means evidence exists: code, a benchmark or result, a design/technical explanation, and a retrospective. Optional depth work raises the ceiling but is never required for the core checkpoint.